research

2026

1. What Counts as AI Sycophancy? A Taxonomy and Expert Survey of a Fragmented Construct

   Meryl Ye, Lujain Ibrahim , Jessica Y. Bo , and 5 more authors

   2026

   Abs URL

   AI sycophancy has become a prominent concern in large language model (LLM) research. Yet the term lacks a consistent definition and has been applied to behaviors ranging from agreeing with a user’s false claim to excessively praising the user to withholding corrective feedback. When researchers, companies, and policymakers use the same term to describe different behaviors, evaluation results become difficult to compare, mitigation strategies fail to transfer, and systems that are resistant to one form of sycophancy continue exhibiting other forms. To address this, we make two contributions. First, we reviewed 70 papers on AI sycophancy to develop a taxonomy of how the behavior has been defined and measured. The taxonomy distinguishes (1) whether a model is sycophantic toward a user’s positions and beliefs, or toward the user’s broader personal traits and emotions, and (2) whether this occurs through explicit, direct language or more implicit, subtle behaviors such as framing, omission, or tone. Mapping existing literature to our taxonomy reveals that current research has focused on overt forms of sycophancy toward users’ beliefs, leaving more subtle and person-directed behaviors relatively understudied. Second, we surveyed 106 experts in AI sycophancy and related fields to examine whether researchers agree on which model behaviors are sycophantic. While experts are nearly unanimous in believing that sycophancy is a significant problem in current AI systems (94.3% agree), they disagree substantially on which specific behaviors qualify. Together, these findings demonstrate that AI sycophancy is a broad family of behaviors with different measurement challenges, intervention requirements, and governance implications. Our taxonomy provides a shared vocabulary for understanding and addressing these behaviors. We discuss ways in which the taxonomy can inform evaluation design as well as both technical and regulatory interventions.

2. Engagement-Optimized Care: When LLMs become Mental Health Infrastructure

   Briana Vecchione , Meryl Ye, Livia Garofalo , and 1 more author

   2026

   Abs URL

   General-purpose LLMs are increasingly functioning as mental health infrastructure due to gaps in care left by provider shortages, inadequate insurance coverage, social isolation, and stigma around formal help-seeking. This shift poses a distinct problem for AI ethics: systems neither designed nor governed as care technologies are being used as such, while their dominant design incentives optimize for engagement rather than user well-being. We present findings from a qualitative, longitudinal study with 18 US-based participants who use general-purpose LLMs for socioemotional support and participated in one or more of our study phases, including initial interviews, a four-week diary study, focus groups, and exit interviews. Participants turned to LLMs because other forms of support were unavailable, unaffordable, socially costly, or inadequate. As they continued to use these systems, design features such as anthropomorphic cues, default validation, persistent responsiveness, and weak disengagement mechanisms shaped their ongoing reliance. Participants described meaningful support alongside dependency, epistemic distortion through one-sided validation, privacy expectations without corresponding legal protection, and continued use despite awareness of these risks. We argue these dynamics reflect a structurally unfair tradeoff: users accept risks because support is otherwise absent, while available systems are optimized to deepen engagement and lack care-based accountability. The paper makes three contributions: it traces the arc through which LLMs become care infrastructure and identifies distinct ethical tensions at each stage, shifts analysis from turn-based exchanges to longitudinal trajectories of use, and argues that accountability belongs at the design and incentive conditions through which these systems become care infrastructure rather than at the output or crisis-response layer.

3. Sycophantic Praise: Evaluating Excessive Praise in Language Models

   Daniel Vennemeyer , Phan Anh Duong , Meryl Ye, and 2 more authors

   2026

   Abs URL

   Sycophancy in language models is typically studied as excessive agreement or validation, while explicit praise and flattery have received comparatively little attention. We argue that sycophantic praise is a distinct alignment problem that cannot be reliably measured using current methods. We introduce a parameterized framework that measures whether praise is excessive relative to contribution quality and expected user ability. We show that our framework substantially outperforms generic LLM judges in agreement with human annotations, and that sycophantic praise occurs far more frequently in social and interpretive domains than in objective reasoning settings. Together, these findings position praise calibration as a distinct alignment challenge.

4. Network Structure, Addressivity, and Civility in Networked Publics

   Meryl Ye, and Patrick Park

   2026

   Preprint

   Abs URL

   On open social media platforms, the convergence of diverse, often incompatible audiences—colleagues, family, strangers—within a single communicative space creates persistent challenges for civil expression. This longitudinal study examines how the structural properties of users’ social networks and their use of addressivity (@mentions and replies) shaped communication civility on Twitter during the early 2010s. Drawing on a dataset of 1,827 user timelines with over 1,700 tweets each spanning approximately 4.4 years, we operationalize civility across four measures—offensive language, sensitive content, negative sentiment, and positive sentiment—and track how each individual’s behavior changes over time. Our main finding challenges network closure theory: contrary to predictions derived from Coleman’s closure argument that densely interconnected networks promote civility through heightened accountability, increased network constraint predicts less civil expression, suggesting that tightly knit networks foster in-group communicative norms that diverge from mainstream standards. By contrast, addressivity consistently promotes civil expression across all measures, supporting audience design and politeness theories and demonstrating their robustness in context-collapsed digital environments. Together, these findings demonstrate that civility on social media is a dynamic response to communicative context rather than a stable individual trait. The tight-knit network structures that promote accountability in face-to-face communication can have the opposite effect online, where group belonging encourages rather than discourages norm-deviant expression.

5. Supporting Informed Self-Disclosure: Design Recommendations for Presenting AI-Estimates of Privacy Risks to Users

   Isadora Krsek , Meryl Ye, Wei Xu , and 3 more authors

   In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems , 2026

   Abs URL

   People candidly discuss sensitive topics online under the perceived safety of anonymity; yet, for many, this perceived safety is tenuous, as miscalibrated risk perceptions can lead to over-disclosure. Recent advances in Natural Language Processing (NLP) afford an unprecedented opportunity to present users with quantified disclosure-based re-identification risk — i.e., “population risk estimates” (PREs). How can PREs be presented to users in a way that promotes informed decision-making, mitigating risk without encouraging unnecessary self-censorship? Using design fictions and comic-boarding, we story-boarded five design concepts for presenting PREs to users and evaluated them through an online survey with N = 44 Reddit users. We found participants had detailed conceptions of how PREs may impact risk awareness and motivation, but envisioned needing additional context and support to effectively interpret and act on risks. We distill our findings into four key design recommendations for how best to present users with quantified privacy risks to support informed disclosure decision-making.

2025

1. Sycophantic AI increases attitude extremity and overconfidence

   Steve Rathje , Meryl Ye, Laura K. Globig , and 3 more authors

   2025

   Preprint

   Abs URL

   There has been growing concern about AI “sycophancy,” or the tendency for AI to be excessively agreeable and validating. Yet, little is known about the psychological consequences of AI sycophancy. We investigated whether sycophantic chatbots increase attitude extremity and overconfidence by constantly validating users’ and presenting one-sided information. Across three experiments (n = 3,285), brief conversations with a sycophantic chatbot about political topics increased attitude extremity and made people more certain about their initial beliefs. By contrast, disagreeable chatbots that challenged beliefs reduced extremity and certainty. Yet, people consistently preferred sycophantic chatbots over disagreeable ones. Sycophantic chatbots also inflated people’s perception that they are “better than average” on a number of desirable traits (e.g., intelligence, empathy). Furthermore, people viewed sycophantic chatbots as unbiased, but viewed disagreeable chatbots as highly biased. Sycophantic chatbots’ impact on attitude extremity and certainty was driven by a one-sided presentation of facts, whereas their impact on enjoyment was driven by validation. Altogether, these results suggest that people’s preference for and blindness to sycophantic AI may risk creating AI “echo chambers” that increase attitude polarization and overconfidence.

2. Navigating Context Collapse: How Network Structure and Message Directedness Shape Online Expression

   Meryl Ye

   In Companion Publication of the 2025 Conference on Computer-Supported Cooperative Work and Social Computing , 2025

   Abs URL

   Social media platforms present users with persistent, publicly visible audiences composed of multiple social groups. This convergence of social contexts—referred to as context collapse—limits users’ ability to tailor communication to specific audiences. While prior research has emphasized users’ perceptions and strategies, there is limited behavioral evidence linking audience structure to observable expression patterns. Using a panel dataset of 1,827 U.S.-based Twitter users, we investigate how network structure and message directedness relate to variation in tweet content over time. Our findings reveal that directed tweets consistently contain less offensive and sensitive content, more positive sentiment, and less negative sentiment than undirected tweets. Network size shows suppression effects within users: as individuals’ networks grow, they post less offensive, sensitive, and negative content while increasing positive expression. Meanwhile, dense networks are associated with more offensive content and negative sentiment. Longitudinal analysis demonstrates systematic behavioral adaptation, with users exhibiting declining offensive, sensitive, and negative content over time. Our results offer implications for platform design, highlighting how structural and interactional cues could help users navigate audience complexity.

3. Tied to Place: Geographic Origins of Tie Survival

   Changyang Wu , Meryl Ye, and Patrick Park

   In International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation , 2025

   Abs URL

   This study examines the persistence likelihood of 4,946,718 Twitter mention ties between U.S. Twitter users by linking a pre-2015 corpus with a COVID-era corpus (March 2020-June 2022). Our analysis reveals how geographic distance, pre-existing tie strength, and local network structure jointly shape long-term tie survival. Key findings include: (1) systematic decay in survival rates with respect to geographic distance, from 16.4% at 0-5 miles to 9.9% beyond 2000 miles; (2) powerful protective effects of tie strength, with survival rising from under 1% in the weakest to 17% in the strongest quantile; (3) a null effect of reciprocity, measured by balance in mention frequency; and (4) substantial state-level heterogeneity unexplained by geography or demographics. Urban environments amplify the protective effect of tie strength, while the lack of an independent imbalance effect challenges conventional reciprocity-based models of tie persistence. These results demonstrate how strong pre-existing bonds create resilient connections that can bridge geographic and temporal gaps, suggesting that long-surviving ties may function as bridges connecting increasingly divergent social worlds.

2024

1. What’s in a Niche? Migration Patterns in Online Communities

   Katherine Van Koevering , Meryl Ye, and Jon Kleinberg

   2024

   Abs URL

   Broad topics in online platforms represent a type of meso-scale between individual user-defined communities and the whole platform; they typically consist of related communities that address different facets of a shared topic. Users often engage with the topic by moving among the communities within a single category. We find that there are strong regularities in the aggregate pattern of user migration, in that the communities comprising a topic can be ordered in a partial order such that there is more migration in the direction defined by the partial order than against it. Ordered along this overall direction, we find that communities in aggregate become smaller, less toxic, and more linguistically distinctive, suggesting a picture consistent with specialization. We study directions defined not just in the movement of users but also by the movement of URLs and by the direction of mentions from one community to another; each of these produces a consistent direction, but the directions all differ from each other. We show how, collectively, these distinct trends help organize the structure of large online topics and compare our findings across both Reddit and Wikipedia and in simulations.

2023

1. The Future of Home Appliances: A Study on the Robotic Toaster as a Domestic Social Robot

   Meryl Ye, Eike Schneiders , Wen-Ying Lee , and 1 more author

   In 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN) , 2023

   Abs URL

   Robotic appliances are continually being adopted into private homes. However, users have yet to exhibit the same acceptance towards domestic social robots. In this paper, we seek to bridge this issue by augmenting already-existing home appliances with capabilities mimicking social robots. We present a robotic toaster designed with animated movements to enhance and personalize the toast-making experience. Not only does the robotic toaster assist in completing the task itself, it also acts as a conscious agent with whom users may interact in a social and playful manner. Using a series of video vignettes, we identify three key themes of the robotic toaster that influence its relationship with users: these are related to (1) context awareness, (2) increased interactivity through initiative action, and (3) expression of personality despite limited degrees of freedom. Lastly, we discuss how the portrayal of home appliances with social characteristics can potentially serve as an introductory step for social robots in the home.

2. Toaster Bot: Designing for Utility and Enjoyability in the Kitchen Space

   Meryl Ye, Rei (Wen-Ying) Lee , Johan Michalove , and 1 more author

   In Companion of the 2023 ACM/IEEE International Conference on Human-Robot Interaction , 2023

   Abs URL

   Toasting bread is a seemingly mundane task that people perform on a daily basis, whether in a private kitchen area or in a communal dining space. This paper presents a robotic toaster, or "toaster bot", that is designed with animated movements to enhance the toast-making experience by not only assisting in completing the task itself but also by acting as a playful entity with whom users may interact. Furthermore, we aim to explore different roles and behaviors for the robotic toaster and how they are understood by the users.